Colors in statistical graphics

What is color?

color
- property possessed by an object of producing different sensations on the eye as a result of the way the object reflects or emits light
- may be difference between what we see and what we perceive
there are three cones that help perceiving different segment of light
- L for long web lenth
- M for medium web length
- S for small web length
in statistical graphics colors are mapped to some variables
- continuous color
- discrete color

How do we see color?

Human eye has
- five million cone cells
- 20 times more rod cells
Cone captures color
Rod captures brightness
Brain processes them and we finally perceive

color-eye

Picture Source: http://jfly.iam.u-tokyo.ac.jp/color/

RGB color space

what is RGB color?
- each color is combination of three primary colors: red, green and blue
- if any two of them are missing, it produces pure third color
- TV and computer screens are equipped with three colors

cols <- rgb(r=c(1,0,0), g=c(0,1,0), b=c(0,0,1))
df <- data.frame(x=1:3, y=1)
ggplot(df, aes(x,y)) + geom_bar(fill=cols, stat="identity")

plot of chunk unnamed-chunk-2

RGB color space

Panels are faceted by color blue

plot of chunk unnamed-chunk-3

HCL Color scheme

problem with RGB
- this is not how human eye perceive color
Our perception of color can be directly controlled by
- hue: color
- chroma: colorness
- luminance: brightness

cols <- hcl(h=c(0,120,240), c=c(90,90,90), l=c(60,60,60))
df <- data.frame(x=1:3, y=1)
ggplot(df, aes(x,y)) + geom_bar(fill=cols, stat="identity")

plot of chunk unnamed-chunk-4

HCL Color scheme

Panels are faceted by luminance
At center chroma is zero
Circular distances represent hue

HSB Color scheme

Our perception of color can be directly controlled by
- hue: color
- saturation: colorness
- luminance: brightness

cols <- hsv(h=c(0,0.33,0.66), s=c(0.9,0.9,0.9), v=c(0.6,0.6,0.6))
df <- data.frame(x=1:3, y=1)
ggplot(df, aes(x,y)) + geom_bar(fill=cols, stat="identity")

plot of chunk unnamed-chunk-6

HSB Color scheme

Panels are faceted by luminance
At center saturation is zero
Circular distances represent hue

plot of chunk unnamed-chunk-7

Comparison of Color schemes

plot of chunk unnamed-chunk-8

Discrete color

Which of these three colors looks more important or emphasized?

plot of chunk unnamed-chunk-9

Discrete color

Which of these three colors looks more important or emphasized?

plot of chunk unnamed-chunk-10

Choosing discrete color

This is a HCL color wheel of luminance = 60
To pick color, choose hue in equal distence
Example : suppose we want to pick 4 colors
- hue : 0, 90, 180, 270
These colors will be equally distent and emphesized
This is how ggplot2 picks default color

plot of chunk unnamed-chunk-11

Choosing discrete color

Evenly spaced hues of equal chroma and luminance. No colour appears more important than any other. Does not imply order.
However, use of more than 7 colors make it harder to distinguish
If you have more than seven categories in the data, perhaphs you need to find an alternative
R uses color as hexadecimal text followed by #
Here we are generating three most distinct colors

discrete-color

hcl(h=c(0,120,240), c=90, l=60)

[1] "#E96485" "#45A500" "#009BE8"

Continuous color

evenly spaced hues between two colours

color vision impairment

vision-impairment

color vision impairment

vision-impairment1

Vision impairment: deuteranopia

Color palette

pie(rep(1, 12), col = rainbow(12))

plot of chunk unnamed-chunk-13

Color palette selection

color-selection

R Color palettes

For discrete data
- brewer
- grey
For continuous data
- gradient2
- gradientn

Reading assignment and references

HCL-Based Color Palettes in R
http://cran.r-project.org/web/packages/colorspace/vignettes/hcl-colors.pdf
Color Design for the Color Vision Impaired by Bernhard Jenny and Nathaniel Vaughn Kelso
http://colororacle.org/resources/2007_JennyKelso_DesigningMapsForTheColourVisionImpaired.pdf
Expert Color Choices for Presenting Data by Maureen Stone, StoneSoup Consulting
http://www.stonesc.com/pubs/Expert%20Color%20Choices.pdf